Use of AI
Of course, you are allowed to use tools like ChatGPT, Elicit, Copilot, or Canva to facilitate your tasks, to have content explained again, or to receive suggestions and inspiration, just as we do.
However, in order to ensure that our work adheres to the principles of good scientific practice and to clarify our expectations regarding the use of these tools, we have compiled a few points below that should be considered.
Responsibility for Content
Everything you submit, send, or publish under your name is your responsibility. This means that you are fully accountable for the content. Specifically, this entails:
You have to at least have a basic understanding of how AI models function (what happens in the background) in order to properly evaluate and interpret the output you receive.
Be aware of potential biases that may affect the answers due to training datasets or algorithms.
Example for bias resulting from training data Only outsource tasks where you can take responsibility for the content, fact-check everything, and have sufficient prior knowledge of the topic
Ethical AI
Potential Reproduction of Biases
This issue arises from the way data and algorithms are used to train AI models. If the data used to train these models is biased or reflects social inequalities, the resulting AI model will also show these biases. This can lead to AI models reinforcing and perpetuating existing prejudices and discriminatory practices without being noticed.
Environmental Impacts
Large-scale AI deploymens are hosted in data centers, which have a significant toll on the planet [@AIHasEnvironmental2024].
For example:
Producing a 2 kg computer requires about 800 kg of raw materials.
Microchips that power AI require rare earth elements, which are often mined in environmentally destructive ways and frequently come from regions affected by civil-wars.
The production of electronics involves materials like lead and mercury, which are harmful to the environment.
Data centers use water during construction and in operation to cool electronic components. Globally, AI-related infrastructure consumes about six times more water than Denmark, which is a problem, considering a quarter of humanity already lacks access to clean water and sanitation.
The use of fossil fuels contributes to the production of greenhouse gases.
A request made through ChatGPT consumes about 10-times the electricity of a Google search.
Reflection questions:
- What is your attitude on the usage of generative AI in class? What are further issues we encounter, when tolerating or not tolerating AI in class?
Copyright
Privacy
Value of human labor
Data exploitation
Shift of power
- What can we do together to ensure an appropriate behavior with AI in class?
- Collect tools we use, so everybody has the same chances, etc.